How will agentic AI redefine growth?
Kartik Venkatesh
Global Head of Innovation
AI has already transformed how businesses operate – enabling greater automation, speed and scalability. But we’re now entering a new phase of innovation: agentic AI. Agentic AI extends predictive models with planning, memory and autonomous action. This shift moves AI from prediction into decision, unlocking new growth opportunities while introducing fresh challenges.
As customer expectations for instant, frictionless onboarding continues to rise, and competition intensifies globally, agentic AI offers a new frontier of innovation. Unlike traditional AI systems that rely on prediction, agentic AI can act independently, adapt in real time and intelligently manage risk. This technology ushes in new ways of working, processes run entirely by well-built agents, can create huge business impact - BCG estimates that effective AI agents can accelerate business processes by 30% to 50%.
But with every leap forward, new vulnerabilities emerge. So how can businesses harness the promise of agentic AI while staying ahead of the risks?
The AI evolution: from GenAI to agentic AI
Three years ago, AI had its “iPhone moment” with the launch of ChatGPT. Generative AI (GenAI) became accessible to the masses, sparking a wave of creativity and automation across industries. Businesses embraced GenAI to generate content at scale, automate customer support, personalise marketing and product recommendations, accelerate internal workflows and much more.
However, while powerful, most enterprise GenAI use cases today are reactive. It responds to prompts, predicts outcomes and supports human decision-making. As a result, most companies have not yet unlocked scalable GenAI value – but those who have are seeing significant returns.
McKinsey reports that while nearly eight in ten companies report using GenAI, just as many report that they haven’t yet seen significant bottom-line impact. They call this the “GenAI paradox.”
The GenAI paradox comes down to a mismatch: enterprise-wide copilots and chatbots have spread rapidly, but their benefits are scattered and tough to measure. Meanwhile, the truly game-changing, specialised uses of AI – where it could transform specific parts of a business – are still stuck in trial mode, with about 90 percent not making it into action.
But now, the tech is going one step further.
McKinsey proposes that Agentic AI is the breakthrough that businesses need to see measurable impact. By acting on its own, planning ahead, remembering what matters and plugging into your systems, it can automate tricky business processes and turn GenAI from a passive helper into a proactive, goal-focused partner that gets things done.
Importantly, agentic AI doesn’t replace GenAI – it builds on it. GenAI is still the core reasoning and language layer that agents rely on. The shift is not from GenAI to agentic AI, but from GenAI-only to GenAI-powered agents.
What is agentic AI?
Agentic AI refers to autonomous systems that can act independently, adapt in real time and intelligently manage risk. These agents don’t just assist – they decide. They reason, learn from outcomes and take action without human intervention.
I recently moderated a panel on this topic at Growth Unlocked with experts from Salesforce, IBM, AWS and GBG (and you can watch the key panel highlights here on demand now). Panelist, Matt Newton – Principal Partner Technical Specialist – AI & Data Platform at IBM, summed up the step change from GenAI to agentic AI simply and perfectly:
“This is the doing phase of AI. We’ve gone from ‘show me’, ‘tell me’, to the ‘do it for me’. And this is the ‘do it for me’ part – a group of agents collaborating with each other with little or no human interaction.”
Missed this year’s Growth Unlocked?
Catch up on the standout moments and key takeaways from our 2025 sessions – all in one place.
Google Cloud defines agentic AI as artificial intelligence built on five core components: perception, reasoning, planning, action and reflection. This put into motion in a continuous cycle allows AI to learn and improve over time. While Salesforce attributes three main features to agentic AI: autonomy, adaptability and goal orientation.
What is key is that agentic AI takes independent action and learns from outcomes. This marks a fundamental shift:
- From task-based to goal-oriented AI
- From static workflows to dynamic decisioning
- From human-directed to self-directed systems
The growth potential that this tech offers businesses is huge. In fact, Gartner forecasts that “by 2028, 15% of day-to-day work decisions will be made autonomously through agentic AI, up from 0% in 2024.”
Though it’s worth highlighting that most agentic systems will be semi-autonomous for the foreseeable future, with humans supervising or approving high-risk decisions.
Agentic commerce: agentic AI in practice
Let’s look at how this technology is already being put into place by organisations and impacting our daily lives as consumers – agentic commerce.
As agentic AI evolves, it’s transforming how people buy, interact and transact online. This emerging trend - agentic commerce - is reshaping buyer behaviour. The idea is that agentic commerce will enable autonomous agents to act on behalf of consumers – shoppers will be able to delegate negotiating, purchasing and fulfilling orders with minimal human input.
What could the future bring? Predictions include digital twins or personal agents will shop, compare and transact on behalf of shoppers and agents will interact with other agents across platforms and businesses, changing the online shopping experience for consumers.
For businesses, this opens new growth opportunities. Agentic commerce is set to have a major impact on retail business operations, offering efficiencies while necessitating strategic adaptations. Early adopters report 6-10% revenue increases and up to 40% improvements in operational efficiency.
In fact, retailers are already rolling out their agentic strategy.
OpenAI announced plans to work on in-agent transactions in 2025 and at the end of September of the same year, Etsy became the first marketplace to support its new checkout experience using the Agentic Commerce Protocol.
Data from 2025’s Black Friday shows how AI is already being used as part of the shopping journey. Shoppers used AI chatbots to hunt for deals and as a result, U.S. retail traffic driven by AI jumped a staggering 805% compared to last year, according to Adobe Analytics. Globally, AI and agents helped influence $14.2 billion in online sales - $3 billion of that from the U.S. alone, according to Salesforce.
The identity industry impact
Looking specifically in the identity space, agentic AI offers transformative potential, including:
- Faster onboarding: Agents can verify identity and assess risk instantly
- Smarter fraud detection: Behavioural analysis and adaptive learning improve accuracy
- Reduced friction: Autonomous systems streamline verification journeys
- Greater inclusivity: Bias reduction and real-time decisioning improve access
But there is also another side of the coin to consider. With autonomy comes complexity – and new vulnerabilities. Agentic AI amplifies existing risks – just as GenAI gave fraudsters greater resources to increase scale and sophistication, so does agentic AI:
- Synthetic identities: Fraudsters can use agents to generate convincing fake profiles
- Adversarial agents: Malicious agents may exploit decisioning systems
- Data poisoning: Compromised training data can skew outcomes
- Agent drift: Autonomous agents may deviate from intended behaviour over time
So, while bringing agentic AI into an organisation is a transformative step, it comes with a unique set of considerations.
Here are some of the critical questions and factors businesses need to keep in mind:
- Identifying and managing drift, bias and rogue behaviour: Are your systems equipped to detect when AI agents diverge from intended behaviour, develop unintended biases, or act outside established boundaries?
- Ensuring explainability and ethics: How will you implement explainable AI and embed ethical design principles to uphold transparency and accountability?
- Strengthening fraud detection: What enhancements can be made to existing fraud tech stacks to recognise new, agent-specific anomalies and behavioural markers of malicious agents?
- Industry collaboration for threat intelligence: In what ways can you collaborate across your industry to share insights and intelligence about emerging threats and attack patterns?
- Preparing for agent-to-agent ecosystems: How will your systems interact securely with external agents, maintain seamless experiences across platforms, and remain compliant in increasingly decentralised networks?
As agentic AI carves out its place within business, the opportunities for accelerated growth and smarter engagement are immense. Yet, this promise comes with a responsibility: organisations must carefully balance risk and opportunity, putting robust safeguards in place to ensure AI acts as a force for progress rather than disruption. As the landscape continues to evolve, staying vigilant, adaptive and ethical will be essential to unlocking sustainable growth.
